arXiv cs.MA· Veronique Ziegler·· 10 天前AI 评分21
当监管者成为干扰源：受控工具调用智能体中控制引发的干扰与成本感知退避
When the Governor Becomes the Disturbance: Control-Generated Disturbance and Cost-Aware Backoff in Governed Tool-Using Agents
AI 导读
研究发现在受控文件恢复环境中，监管强度提升会触发工具故障，成本盲目的监管者会将故障转化为持续阻塞导致任务无法完成。采用基于已知诱发事件移动平均的自适应退避规则可降低干预概率，在 Gemini 2.5 Flash 的 576 个回合、6 项任务实验中减少了阻塞并提升完成率，中等退避强度取得最高聚合成功率。
正文 · 原文
Abstract:Supervisory governors can interfere with the tool-using agents they regulate. We study this possibility in a controlled file-recovery environment where increases in regulatory intensity trigger experimentally imposed tool failures. A cost-blind governor can turn these failures into persistent blocking that prevents task completion. We compare this governor with a backoff rule that reduces intervention probability using a moving average of known induced events. On a hand-coded stochastic-policy agent, the failure pattern appears under both result replacement and execution of corrupted tool arguments. For the persistent policy, adaptive backoff improves completion relative to a fixed weak governor with approximately matched intervention frequency. A Gemini 2.5 Flash experiment comprising 576 episodes across 6 tasks also shows reduced blocking and improved completion under backoff; among the tested settings, intermediate backoff strength achieves the highest observed aggregate success. These results identify an interaction between intervention cost and persistent action blocking, together with a possible mitigation. The cost mechanisms are imposed and their induced events are directly observable to the backoff rule; applicability beyond this controlled environment remains an empirical question.
| Comments: | Includes ancillary experimental data and a Python script for reproducing numerical summaries. Code: this https URL |
| Subjects: | Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.09037 [cs.AI] |
| (or arXiv:2610.09037v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.09037 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Veronique Ziegler [view email]
[v1]
Tue, 6 Oct 2026 19:39:24 UTC (27 KB)
来源：arXiv cs.MA · arxiv.org